Speech recognition with a cochlear implant using triphasic charge-balanced pulses

Abstract

ObjectiveTypically, symmetrical charge-balanced biphasic current pulses are used in cochlear implants to ensure biological safety. Theoretically, monophasic pulses are more effective, but potentially noxious, stimuli. In this study we charge-balanced such monophasic pulses during selected non-stimulated intervals, effectively leading to triphasic pulses with a 4:1 amplitude ratio between the cathodic and anodic phases. Apart from ensuring safety, this is also expected to reduce power consumption and channel interaction. Material and MethodsSeven experienced Clarion CII cochlear implant users with a multichannel (12–16 channels) monopolar continuous interleaved sampling (CIS) strategy participated in the study. Three different CIS strategies were fitted using the Clarion Research Interface (CRI-2). The reference was an implementation of each subject's own CIS program. The two strategies tested used triphasic pulses on the same channels, one with half-wave rectification (TP-HWR) and one without rectification (TP-NoR) at the input. Directly after fitting (i.e. without any training), speech perception (phoneme score on consonant–vowel–consonant words) was measured in silence (sound-only) and in speech-shaped background noise with signal:noise ratios (SNRs) of +5 and 0 dB. ResultsSpeech perception with the reference via the CRI-2 was equal to that of the free-field condition with the subjects’ own speech processor. With the TP-NoR strategy, speech perception improved significantly (from 89% to 93%) in silence and in the 0-dB SNR condition (from 43% to 49%). With a SNR of +5 dB, performance was stable at ≈66%. With the TP-HWR strategy, performance increased significantly in the 0- and +5-dB SNR conditions, to 55% and 74%, respectively. Power consumption was reduced in both strategies, to 30% and 36% for TP-HWR and TP-NoR, respectively. Conclusion The new triphasic strategies are most promising, with respect to both their improved speech perception and reduced power requirements. The optimal parameters will have to be identified following long-term use.